Introduction

Claude Code is Anthropic's agentic coding tool that lives in your terminal, IDE, and browser. It reads your codebase, edits files, runs commands, and integrates with development tools through MCP servers and plugins. With dozens of built-in slash commands, CLI flags, bundled skills, and keyboard shortcuts, it can be hard to remember everything that is available.

This cheat sheet compiles the most useful Claude Code commands into one color-coded, printable reference that you can keep next to your terminal. Hit the print button below to get a clean A3 landscape version, or scroll through the full reference on this page.

Session Management

Session commands control the lifecycle of your conversation with Claude Code. The most important distinction to understand is between /compact and /clear. Compacting summarizes the conversation to reclaim token space while preserving the thread of what was discussed. Clearing wipes everything and starts fresh. As a rule of thumb, compact when your context usage approaches 80%, and clear when you are switching to a completely unrelated task.

The /branch command is useful when you want to explore an alternative approach without losing your current conversation state. Think of it as a git branch for your Claude Code session. And /rewind lets you roll back both the conversation and any code changes to an earlier checkpoint.

Context & Information

Understanding your context usage is key to efficient Claude Code sessions. The /context command gives you a visual grid showing how much of the context window is consumed and by what. If you are on a paid plan, /cost breaks down token consumption and estimated costs for the current session.

The /diff command opens an interactive viewer that shows both your uncommitted git changes and the changes Claude made in each turn, so you can review exactly what happened.

Model & Mode Control

Claude Code supports multiple models and effort levels. Use /model to switch between them mid-session. The arrow keys in the model picker let you adjust the effort level simultaneously. The effort setting controls how deeply Claude reasons: low is fast and cheap for mechanical tasks, high gives thorough reasoning, and max unlocks the deepest thinking on Opus 4.6 (current session only).

/fast enables speed-optimized API settings. It is not a different model. It runs the same Opus 4.6 with faster parameters, making it ideal for rapid interactive iteration. Turn it off when cost efficiency matters more than speed.

/plan puts Claude into a read-only reasoning mode where it analyzes and plans without making changes. This is helpful for understanding complex codebases before committing to edits.

Configuration & Permissions

The /config command opens a settings interface where you can adjust theme, model preferences, output style, and editor mode. For controlling what Claude is allowed to do, /permissions lets you manage allow, ask, and deny rules at different scopes.

/memory gives you access to CLAUDE.md files and the auto-memory system. Claude builds auto-memory as it works, saving learnings like build commands and debugging insights across sessions. You can also manually edit your project's CLAUDE.md to set coding standards, architecture decisions, and conventions that Claude should follow.

The /hooks command shows your lifecycle hooks, which are shell commands that run before or after Claude Code actions. For example, you could auto-run Prettier after every file edit or execute a type check before accepting changes.

Extensions & Integrations

Claude Code's extensibility comes from MCP servers, plugins, and native integrations. The /mcp command manages your Model Context Protocol server connections, which expose external tools and data sources directly inside your session. Connected MCP servers contribute their own slash commands in the /mcp__<server>__<prompt> format.

Plugins extend Claude Code with skills, agents, hooks, and even LSP servers. Use /plugin to discover and install them from marketplaces, and /reload-plugins to hot-reload after changes without restarting.

The GitHub and Slack integrations (/install-github-app and /install-slack-app) connect Claude to your team workflows. With the GitHub app, Claude can participate in pull request reviews via GitHub Actions.

Bundled Skills

Claude Code ships with several bundled skills that appear alongside built-in commands when you type /. The standout is /simplify, which replaced the older /review command and runs a three-agent parallel code review. /batch lets you run a task across multiple files, /debug provides a structured debugging workflow, and /security-review scans the current branch diff for vulnerabilities.

For quick questions that do not need to consume context, use /btw. It fires off a side question without adding to the main conversation, keeping your context clean.

The /loop skill is particularly useful for monitoring. Give it an interval and a prompt, and it creates a cron job that fires in the background: /loop 5m check if the deployment finished. Supported units are s (seconds, rounded to nearest minute), m, h, and d. These are session-scoped and stop when you exit.

CLI Flags

The -p (or --print) flag runs Claude Code non-interactively, making it scriptable for CI, cron jobs, and automation. Combined with --output-format json and --json-schema, you can get structured, machine-readable output.

For unattended execution, --dangerously-skip-permissions skips all approval prompts. Use it with care and scope your prompts tightly. The --allowedTools flag is a more granular alternative, letting you pre-approve specific tools like Bash, Read, and Edit.

System prompt customization is available through four flags. --append-system-prompt and --system-prompt-file are the most commonly used. Appending preserves Claude Code's built-in capabilities while adding your instructions, which is the right choice for most use cases.

The --bare flag is useful for CI scripts. It skips auto-discovery of hooks, skills, plugins, MCP servers, auto-memory, and CLAUDE.md, ensuring consistent behavior across machines.

Keyboard Shortcuts

Shift+Tab is perhaps the most useful shortcut, cycling between normal mode, auto-accept mode, and plan mode. Ctrl+O toggles verbose output so you can see Claude's extended thinking. And pressing Ctrl+F twice rapidly kills all background agents, which is a lifesaver when something gets stuck.

The @ prefix lets you attach files or directories to the conversation context without waiting for Claude to read them. And the ! prefix runs a shell command directly with conversation context, so ! git status runs in your shell and the output becomes part of the conversation.

Scheduling Prompts

Claude Code supports multiple scheduling approaches. For in-session recurring work, use /loop or /schedule (which creates cloud tasks on Anthropic infrastructure). For persistent unattended execution on your own server, the at command or cron combined with claude -p works well:

Make sure to use the full path to the claude binary, cd into your project directory first (so Claude Code picks up your CLAUDE.md and project context), and redirect both stdout and stderr to capture the full output.

Creating Custom Skills

Beyond the built-in commands, you can create your own slash commands as skills. Create a directory under .claude/skills/ in your project (for team-shared skills) or ~/.claude/skills/ (for personal skills), add a SKILL.md file with YAML frontmatter and markdown instructions, and your new command appears in the / menu automatically.

Skills support frontmatter fields like allowed-tools (to grant tool access without per-use approval), model (to override the model for that skill), context: fork (to run in a sub-agent), and disable-model-invocation: true (to prevent Claude from auto-invoking the skill). For detailed guidance, see the Skills documentation.